How to use create_activation method in localstack

Best Python code snippet using localstack_python

layers_factory.py

Source:layers_factory.py

...65        input_dim=input_dim,66        output_dim=int(layer_info['output_dim']),67        dropout=parse_as_bool(layer_info['dropout']),68        sparse_inputs=parse_as_bool(layer_info['sparse_inputs']),69        act=create_activation(layer_info['act']),70        bias=parse_as_bool(layer_info['bias']),71        featureless=False,72        num_supports=1)73def create_GraphConvolutionAttention_layer(layer_info, model, layer_id):74    if not 5 <= len(layer_info) <= 6:75        raise RuntimeError('GraphConvolution layer must have 3-4 specs')76    input_dim = layer_info.get('input_dim')77    if not input_dim:78        if layer_id != 1:79            raise RuntimeError(80                'The input dim for layer must be specified'.format(layer_id))81        input_dim = model.input_dim82    else:83        input_dim = int(input_dim)84    return GraphConvolutionAttention(85        input_dim=input_dim,86        output_dim=int(layer_info['output_dim']),87        dropout=parse_as_bool(layer_info['dropout']),88        sparse_inputs=parse_as_bool(layer_info['sparse_inputs']),89        act=create_activation(layer_info['act']),90        bias=parse_as_bool(layer_info['bias']),91        featureless=False,92        num_supports=1)93def create_Coarsening_layer(layer_info):94    if not len(layer_info) == 1:95        raise RuntimeError('Coarsening layer must have 1 spec')96    return Coarsening(pool_style=layer_info['pool_style'])97def create_Average_layer(layer_info):98    if not len(layer_info) == 0:99        raise RuntimeError('Average layer must have 0 specs')100    return Average()101def create_Attention_layer(layer_info):102    if not len(layer_info) == 5:103        raise RuntimeError('Attention layer must have 5 specs')104    return Attention(input_dim=int(layer_info['input_dim']),105                     att_times=int(layer_info['att_times']),106                     att_num=int(layer_info['att_num']),107                     att_style=layer_info['att_style'],108                     att_weight=parse_as_bool(layer_info['att_weight']))109def create_Dot_layer(layer_info):110    if not len(layer_info) == 1:111        raise RuntimeError('Dot layer must have 1 specs')112    return Dot(output_dim=int(layer_info['output_dim']))113def create_SLM_layer(layer_info):114    if not len(layer_info) == 5:115        raise RuntimeError('SLM layer must have 5 specs')116    return SLM(117        input_dim=int(layer_info['input_dim']),118        output_dim=int(layer_info['output_dim']),119        act=create_activation(layer_info['act']),120        dropout=parse_as_bool(layer_info['dropout']),121        bias=parse_as_bool(layer_info['bias']))122def create_NTN_layer(layer_info):123    if not len(layer_info) == 6:124        raise RuntimeError('NTN layer must have 6 specs')125    return NTN(126        input_dim=int(layer_info['input_dim']),127        feature_map_dim=int(layer_info['feature_map_dim']),128        dropout=parse_as_bool(layer_info['dropout']),129        inneract=create_activation(layer_info['inneract']),130        apply_u=parse_as_bool(layer_info['apply_u']),131        bias=parse_as_bool(layer_info['bias']))132def create_ANPM_layer(layer_info):133    if not len(layer_info) == 14:134        raise RuntimeError('ANPM layer must have 14 specs')135    return ANPM(136        input_dim=int(layer_info['input_dim']),137        att_times=int(layer_info['att_times']),138        att_num=int(layer_info['att_num']),139        att_style=layer_info['att_style'],140        att_weight=parse_as_bool(layer_info['att_weight']),141        feature_map_dim=int(layer_info['feature_map_dim']),142        dropout=parse_as_bool(layer_info['dropout']),143        bias=parse_as_bool(layer_info['bias']),144        ntn_inneract=create_activation(layer_info['ntn_inneract']),145        apply_u=parse_as_bool(layer_info['apply_u']),146        padding_value=int(layer_info['padding_value']),147        mne_inneract=create_activation(layer_info['mne_inneract']),148        # num_bins=int(layer_info['num_bins'])149        mne_method=layer_info['mne_method'],150        branch_style=layer_info['branch_style'])151def create_ANPMD_layer(layer_info):152    if not len(layer_info) == 22:153        raise RuntimeError('ANPMD layer must have 22 specs')154    return ANPMD(155        input_dim=int(layer_info['input_dim']),156        att_times=int(layer_info['att_times']),157        att_num=int(layer_info['att_num']),158        att_style=layer_info['att_style'],159        att_weight=parse_as_bool(layer_info['att_weight']),160        feature_map_dim=int(layer_info['feature_map_dim']),161        dropout=parse_as_bool(layer_info['dropout']),162        bias=parse_as_bool(layer_info['bias']),163        ntn_inneract=create_activation(layer_info['ntn_inneract']),164        apply_u=parse_as_bool(layer_info['apply_u']),165        padding_value=int(layer_info['padding_value']),166        mne_inneract=create_activation(layer_info['mne_inneract']),167        mne_method=layer_info['mne_method'],168        branch_style=layer_info['branch_style'],169        dense1_dropout=parse_as_bool(layer_info['dense1_dropout']),170        dense1_act=create_activation(layer_info['dense1_act']),171        dense1_bias=parse_as_bool(layer_info['dense1_bias']),172        dense1_output_dim=int(layer_info['dense1_output_dim']),173        dense2_dropout=parse_as_bool(layer_info['dense2_dropout']),174        dense2_act=create_activation(layer_info['dense2_act']),175        dense2_bias=parse_as_bool(layer_info['dense2_bias']),176        dense2_output_dim=int(layer_info['dense2_output_dim']))177def create_ANNH_layer(layer_info):178    if not len(layer_info) == 14:179        raise RuntimeError('ANNH layer must have 14 specs')180    return ANNH(181        input_dim=int(layer_info['input_dim']),182        att_times=int(layer_info['att_times']),183        att_num=int(layer_info['att_num']),184        att_style=layer_info['att_style'],185        att_weight=parse_as_bool(layer_info['att_weight']),186        feature_map_dim=int(layer_info['feature_map_dim']),187        dropout=parse_as_bool(layer_info['dropout']),188        bias=parse_as_bool(layer_info['bias']),189        ntn_inneract=create_activation(layer_info['ntn_inneract']),190        apply_u=parse_as_bool(layer_info['apply_u']),191        padding_value=int(layer_info['padding_value']),192        mne_inneract=create_activation(layer_info['mne_inneract']),193        # num_bins=int(layer_info['num_bins'])194        mne_method=layer_info['mne_method'],195        branch_style=layer_info['branch_style'])196def create_Dense_layer(layer_info):197    if not len(layer_info) == 5:198        raise RuntimeError('Dense layer must have 5 specs')199    return Dense(200        input_dim=int(layer_info['input_dim']),201        output_dim=int(layer_info['output_dim']),202        dropout=parse_as_bool(layer_info['dropout']),203        act=create_activation(layer_info['act']),204        bias=parse_as_bool(layer_info['bias']))205def create_Padding_layer(layer_info):206    if not len(layer_info) == 1:207        raise RuntimeError('Padding layer must have 1 specs')208    return Padding(209        padding_value=int(layer_info['padding_value']))210def create_MNE_layer(layer_info):211    if not len(layer_info) == 3:212        raise RuntimeError('MNE layer must have 3 specs')213    return MNE(214        input_dim=int(layer_info['input_dim']),215        dropout=parse_as_bool(layer_info['dropout']),216        inneract=create_activation(layer_info['inneract']))217def create_CNN_layer(layer_info):218    if not len(layer_info) == 11:219        raise RuntimeError('CNN layer must have 9 specs')220    return CNN(221        start_cnn=parse_as_bool(layer_info['start_cnn']),222        end_cnn=parse_as_bool(layer_info['end_cnn']),223        window_size=int(layer_info['window_size']),224        kernel_stride=int(layer_info['kernel_stride']),225        in_channel=int(layer_info['in_channel']),226        out_channel=int(layer_info['out_channel']),227        padding=layer_info['padding'],228        pool_size=int(layer_info['pool_size']),229        dropout=parse_as_bool(layer_info['dropout']),230        act=create_activation(layer_info['act']),231        bias=parse_as_bool(layer_info['bias']))232def create_activation(act, sim_kernel=None, use_tf=True):233    if act == 'relu':234        return tf.nn.relu if use_tf else relu_np235    elif act == 'identity':236        return tf.identity if use_tf else identity_np237    elif act == 'sigmoid':238        return tf.sigmoid if use_tf else sigmoid_np239    elif act == 'tanh':240        return tf.tanh if use_tf else np.tanh241    elif act == 'sim_kernel':242        return sim_kernel.dist_to_sim_tf if use_tf else \243            sim_kernel.dist_to_sim_np244    else:245        raise RuntimeError('Unknown activation function {}'.format(act))246def relu_np(x):...

MRA_Model.py

Source:MRA_Model.py

...4import numpy as np5import cv26from tensorflow.keras import models,layers7from wavetf import WaveTFFactory8def create_activation():9    return layers.ReLU()10def sigmoid_activation():11    return layers.Softmax()12def create_model():13    input_1 = layers.Input(batch_input_shape=(4,128,128,3))14    enc_1 = layers.Conv2D(64,kernel_size=(3,3),padding="same",kernel_initializer = tf.random_normal_initializer(stddev=0.01))(input_1)15    enc_1 = create_activation()(enc_1)16    enc_1 = layers.Conv2D(64,kernel_size=(3,3),padding="same",kernel_initializer = tf.random_normal_initializer(stddev=0.01))(enc_1)17    enc_1 = create_activation()(enc_1)    #This will be used in skip connection 118    maxpool_1 = layers.MaxPooling2D((2,2))(enc_1)19    ######################################20    '''21    Take the DL1 block here and concatenate with the downstream layer22    '''23    w = WaveTFFactory().build('db2', dim=2)24    dl1_input = w.call(tf.expand_dims(input_1[:,:,:,0],axis=-1)) #Calculate Wavelet Decomposition on only gray scale image25    dl1_enc = layers.Conv2D(64,kernel_size=(3,3),padding="same",kernel_initializer = tf.random_normal_initializer(stddev=0.01))(dl1_input)26    dl1_enc = create_activation()(dl1_enc)27    dl1_enc = layers.Conv2D(64,kernel_size=(3,3),padding="same",kernel_initializer = tf.random_normal_initializer(stddev=0.01))(dl1_enc)28    dl1_enc = create_activation()(dl1_enc)29    #Concatenate with main mode i.e output of maxpool_130    concat_1 = tf.concat([maxpool_1,dl1_enc],axis=-1)31    enc_2 = layers.Conv2D(128,kernel_size=(3,3),padding="same",kernel_initializer = tf.random_normal_initializer(stddev=0.01))(concat_1)32    enc_2 = create_activation()(enc_2)33    enc_2 = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(enc_2)34    enc_2 = create_activation()(enc_2) #This will be used in skip connection 235    maxpool_2 = layers.MaxPooling2D((2,2))(enc_2)36    ######################################37    '''38    Take the DL2 block here and concatenate with the downstream layer39    '''40    dl2_input = w.call(tf.expand_dims(dl1_input[:, :, :, 0], axis=-1))  # Calculate Wavelet Decomposition on only channel 0 of dl1_input41    dl2_enc = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dl2_input)42    dl2_enc = create_activation()(dl2_enc)43    dl2_enc = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dl2_enc)44    dl2_enc = create_activation()(dl2_enc)45    concat_2 = tf.concat([maxpool_2,dl2_enc],axis=-1)46    enc_3 = layers.Conv2D(256, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(concat_2)47    enc_3 = create_activation()(enc_3)48    enc_3 = layers.Conv2D(256, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(enc_3)49    enc_3 = create_activation()(enc_3)  # This will be used in skip connection 350    maxpool_3 = layers.MaxPooling2D((2, 2))(enc_3)51    enc_4 = layers.Conv2D(512, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(maxpool_3)52    enc_4 = create_activation()(enc_4)53    enc_4 = layers.Conv2D(512, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(enc_4)54    enc_4 = create_activation()(enc_4)  # This will be used in skip connection 455    maxpool_4 = layers.MaxPooling2D((2, 2))(enc_4)56    enc_5 = layers.Conv2D(1024, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(maxpool_4)57    enc_5 = create_activation()(enc_5)58    enc_5 = layers.Conv2D(1024, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(enc_5)59    enc_5 = create_activation()(enc_5)60    #####-> Downsampling Layer Ends, from now onwards it is upsampling side61    upsampling_1 = layers.UpSampling2D((2,2))(enc_5)62    dec_1 = layers.Conv2D(512,kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(upsampling_1)63    dec_1 = create_activation()(dec_1)64    concat_3 = tf.concat([enc_4,dec_1],axis = -1)65    dec_2 = layers.Conv2D(512, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(concat_3)66    dec_2 = create_activation()(dec_2)67    dec_2 = layers.Conv2D(512, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_2)68    dec_2 = create_activation()(dec_2)69    upsampling_2 = layers.UpSampling2D((2,2))(dec_2)70    dec_3 = layers.Conv2D(256, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(upsampling_2)71    dec_3 = create_activation()(dec_3)72    concat_4 = tf.concat([enc_3,dec_3],axis=-1)73    dec_4 = layers.Conv2D(256, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(concat_4)74    dec_4 = create_activation()(dec_4)75    dec_4 = layers.Conv2D(256, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_4)76    dec_4 = create_activation()(dec_4)77    upsampling_3 = layers.UpSampling2D((2, 2))(dec_4)78    dec_5 = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(upsampling_3)79    dec_5 = create_activation()(dec_5)80    concat_5 = tf.concat([enc_2,dec_5],axis=-1)81    dec_6 = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(concat_5)82    dec_6 = create_activation()(dec_6)83    dec_6 = layers.Conv2D(128, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_6)84    dec_6 = create_activation()(dec_6)85    upsampling_4 = layers.UpSampling2D((2, 2))(dec_6)86    dec_7 = layers.Conv2D(64, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(upsampling_4)87    dec_7 = create_activation()(dec_7)88    concat_6 = tf.concat([enc_1, dec_7], axis=-1)89    dec_8 = layers.Conv2D(64, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(concat_6)90    dec_8 = create_activation()(dec_8)91    dec_8 = layers.Conv2D(64, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_8)92    dec_8 = create_activation()(dec_8)93    dec_9 = layers.Conv2D(4, kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_8)94    dec_9 = create_activation()(dec_9)95    final_output = layers.Conv2D(4 , kernel_size=(3, 3), padding="same", kernel_initializer=tf.random_normal_initializer(stddev=0.01))(dec_9)96    #final_output = sigmoid_activation()(final_output)97    #model = models.Model(input_1,final_output)...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.